Systemd Socket Summary / socket failed (check_mk)

[yetanotheraccount]

Hello all.

After a few hours of reading forum posts + try&error, I decided to join the forum and ask the experts for help.

I am running CheckMK RAW 2.2.0.p10 on an Ubuntu 22.04.3 server.

I am monitoring eight (8) Debian 11 servers + the CheckMK server itself. CheckMK agents 2.2.0p10 are installed on all servers.

Everything is running fine (I think) except for these annoying messages I get on 5 of the 8 servers.
The messages appeared after upgrading the agents to 2.2.0p10.

I already tried to compare configs, etc. from a server with the problems to a server without problems. No luck…

Maybe someone can point me in the right direction?

[lkoenig (Lorenz Koenig)]

What do you see when you check the status of the socket-unit on these systems (systemctl status check_mk.socket)?

[yetanotheraccount]

Then I see this:

root@my2:~# systemctl status check_mk.socket
● check_mk.socket
Loaded: not-found (Reason: Unit check_mk.socket not found.)
Active: failed (Result: resources) since Fri 2023-09-22 10:15:38 CEST; 1 day 3h ago
CPU: 665ms

Sep 21 11:22:50 my2 systemd[1]: Listening on Check_MK Agent Socket.
Sep 22 10:15:38 my2 systemd[1]: check_mk.socket: Failed to queue service startup job (Maybe the service file is missing or not a template unit?): Invalid argument
Sep 22 10:15:38 my2. systemd[1]: check_mk.socket: Failed with result ‘resources’.

I know it shows “failed” - but I can´t figure out why…

[lkoenig (Lorenz Koenig)]

I know with some update, i can’t tell which one it was, there was an update to the systemd-units that shipped with the agent and there the names were changed as well. At the least with the introduction of the cmk-agent-ctl this must have happened.
So it looks like with the update the check_mk.socket unit somehow still lingers around in your systemd-configuration, event though the socket-file does not exist anymore - as seen in the output Unit check_mk.socket not found
I would try the following: Check if there is a unit cmk-agent-ctl-daemon.service. The process started by this unit should be the one listening on port 6556 now instead of the “old” check_mk.socket. Then i would reload the systemd-configuration (systemctl daemon-reload) and if the only unit that is still marked as failed is the obsolete check_mk.socket i would just run systemctl reset-failed.

[yetanotheraccount]

You made my day !

systemctl reset-failed fixed it !

Thank you very much!